Emotion classification of artistic images using domain adaptation and transfer learning

As society evolves, the appreciation and pursuit of art continue to grow. However, current technology struggles to intelligently interpret the emotional expressions conveyed in images. To enhance the understanding of emotions expressed in artistic images, we propose a novel emotion classification method that integrates domain adaptation and transfer learning. We first introduce an attention-based salient feature extraction technique designed to emphasize the primary artistic elements within an image and enhance the corresponding regions. Leveraging these salient features, we then develop a domain-adaptive image emotion classification model to capture semantic information and accurately recognize the emotional essence of artistic content. Experimental results validate the effectiveness of our approach, achieving a mean average precision (mAP) of 92.4% and an accuracy of 98.9%, demonstrating its capability to provide precise emotional interpretations of artworks. Our method offers a significant advancement in the intelligent analysis of artistic images, combining attention mechanisms, domain adaptation, and transfer learning to improve emotional understanding in visual art.